topic Re: VMS 8-3, IA64 crash assistance in Operating System - OpenVMS

VMS 8-3, IA64 crash assistance

Brian Reiter — Wed, 21 Mar 2007 04:56:29 GMT

Hi there,

A multi-threaded process (written in Pascal) seems to be the main cause of a system crash (some details attached). Its been sometime since I've looked at crash dumps (alpba) so I was wondering if there was any information/tips specific to the Itanium platform?

cheers

Brian

Re: VMS 8-3, IA64 crash assistance

Volker Halle — Wed, 21 Mar 2007 05:39:15 GMT

Brian,

the system is crashing due to an ASTFLT exception.

Make sure you've installed the most recent patches. VMS83I_UPDATE-V0200 and a couple of others have been released during the last 4 weeks.

Volker.

Re: VMS 8-3, IA64 crash assistance

Brian Reiter — Wed, 21 Mar 2007 05:44:12 GMT

Just installed in the new update patch. Could be sometime before I know if its helped or not.

Re: VMS 8-3, IA64 crash assistance

Robert Gezelter — Wed, 21 Mar 2007 06:03:56 GMT

Brian,

The general techniques of interpreting a dump are similar to those on Alpha, albeit with a different instruction set and hardware registers.

Even though you have applied the patch, I would ensure that the dump is archived, in the event that the problem recurs. If the problem recurs, I would recommend escalating the problem to the support center.

- Bob Gezelter, http://www.rlgsc.com

Re: VMS 8-3, IA64 crash assistance

Andreas Gruhl — Wed, 21 Mar 2007 07:58:17 GMT

There is a bug in PASCAL 5.9 / I64, which may cause the user stack pointer to temporarily become negative. This is only for the duration of a few instructions, but when at this very moment an AST arrives VMS crashes. This bug is fixed in PASCAL V6.0 (which I think is not in the official distribution yet).

There is also a fix in VMS causing the system to survive and only the process to be killed.
It involves a modified SYS$LOADABLE_IMAGES:EXCEPTION.EXE (Link File Id X-5, Link Time 18-SEP-2006). The HP internal reference is CLDNOTES #6027. I haven't seen a public available patch yet.

Re: VMS 8-3, IA64 crash assistance

Brian Reiter — Wed, 21 Mar 2007 08:16:45 GMT

Got Pascal V6 already (had other issues with the Itanium version which it fixed) so hopefully that is out.

The current version of Exception.exe has a link time of 29-JUN-2006 23:20:22.77, is the implication then that I've missed a patch somewhere?

cheers

Brian

Re: VMS 8-3, IA64 crash assistance

Volker Halle — Wed, 21 Mar 2007 08:22:51 GMT

Brian,

please carefully check, whether ALL PASCAL modules in your application have been re-complied with PASCAL V6 !

As far as I can see, there have been no patches issued for OpenVMS I64 V8.3, which have included EXCEPTION.EXE. Andreas has received a special version from HP...

So if all of your PASCAL modules have been recompiled with PASCAL V6, consider to escalate this crash problem to HP.

Volker.

Re: VMS 8-3, IA64 crash assistance

Volker Halle — Wed, 21 Mar 2007 08:47:26 GMT

Brian,

once you know what you need to be looking for...

USP = FFFFFFFF.FFFEFD00

This does not look like a healthy value for the user mode stack pointer.

Volker.

Re: VMS 8-3, IA64 crash assistance

Brian Reiter — Wed, 21 Mar 2007 08:55:58 GMT

Volker,

Looks like all the code has been compiled with the right version of the Pascal compiler. Installation of the patch made no difference as the system has just gone again (the periods betweem crashes is random).

The USP does look wrong.

I've asked our suppliers to raise a call with HP.

cheers

Brian

Re: VMS 8-3, IA64 crash assistance

Andreas Gruhl — Wed, 21 Mar 2007 10:34:13 GMT

Brian,

there are different flavors of PASCAL around.
Our (corrected) version is V6.0-110.
I think V6.0-107 did still contain the bug.
Best ask JR if you got a version below 110.

Andreas

Re: VMS 8-3, IA64 crash assistance

Brian Reiter — Wed, 21 Mar 2007 10:55:59 GMT

Andreas,

Ah. Just my luck then :) Pascal compiler claims to be 107.

Ah well

cheers

Brian

Re: VMS 8-3, IA64 crash assistance

John Reagan — Thu, 22 Mar 2007 08:47:04 GMT

Andreas is correct. The fix was in V6.0-110 and beyond. I'll contact Brian offline for access to a fixed compiler. There isn't an ECO to Pascal just yet. It is on my list to do.

Re: VMS 8-3, IA64 crash assistance

Brian Reiter — Fri, 23 Mar 2007 03:03:29 GMT

The new compiler version seems to have resolved the OpenVMS crash. I now have other problems with infinite loops in previously well behaved processes, but that is another story and another fun day ahead.

thanks for your help

Brian